Background: MR diffusion-weighted imaging (DWI) shows acute ischemic lesions early after stroke so it might improve outcome prediction and reduce sample sizes in stroke treatment trials. Previous studies of DWI and outcome produced conflicting results. Objective: To determine whether DWI lesion characteristics independently predict outcome in a broad range of patients with acute stroke. Methods: The authors recruited hospital-admitted patients with all severities of suspected stroke, assessed stroke severity on the NIH Stroke Scale (NIHSS), performed early brain DWI, and assessed outcome at 3 months (modified Rankin Scale). Clinical data and DWI lesion parameters were evaluated in a logistic regression model to identify independent predictors of outcome at 3 months and a previously described “Three-Item Scale” (including DWI) was tested for outcome prediction. Results: Among 82 patients (mean NIHSS 7.1 [+/- 6.3 SD]), the only independent outcome predictors were age and stroke severity. Neither DWI lesion volume nor apparent diffusion coefficient nor the previously described Three-Item Scale predicted outcome independently. Comparison with previous studies suggested that DWI may predict outcome only in patients with more severe cortical ischemic strokes. Conclusions: Across a broad range of stroke severities, diffusion-weighted imaging (DWI) did not predict outcome beyond that of key clinical variables. Thus, DWI is unlikely to reduce sample sizes in acute stroke trials assessing functional outcome, especially where estimated treatment effects are modest.